Systems Biology Approaches to Study the Molecular Effects of Caloric Restriction and Polyphenols on Aging Processes

Genes Nutr (2015) 10:58 DOI 10.1007/s12263-015-0508-9

RESEARCH PAPER

Systems biology approaches to study the molecular effects of caloric restriction and polyphenols on aging processes

1 1 1 1 Se´bastien Lacroix • Mario Lauria • Marie-Pier Scott-Boyer • Luca Marchetti • 1,2 1 Corrado Priami • Laura Caberlotto

Received: 17 February 2015 / Accepted: 13 November 2015 / Published online: 25 November 2015 Ó Springer-Verlag Berlin Heidelberg 2015

Abstract Worldwide population is aging, and a large part some examples on how to extend the application of these of the growing burden associated with age-related condi- methods using available microarray data. tions can be prevented or delayed by promoting healthy lifestyle and normalizing metabolic risk factors. However, Keywords Network analysis Á Multi-omics integration Á a better understanding of the pleiotropic effects of available Glutamate receptors Á Oxidative stress Á Inflammation Á nutritional interventions and their influence on the multiple Healthy aging processes affected by aging is needed to select and implement the most promising actions. New methods of analysis are required to tackle the complexity of the The molecular basis of nutritional strategies interplay between nutritional interventions and aging, and to promote healthy aging to make sense of a growing amount of -omics data being produced for this purpose. In this paper, we review how Global life expectancy and proportion of people aged over various systems biology-inspired methods of analysis can 60 years are increasing. Aging-associated comorbidities be applied to the study of the molecular basis of nutritional will become the next global public health challenge in the interventions promoting healthy aging, notably caloric context of westernization of daily habits and rising preva- restriction and polyphenol supplementation. We specifi- lence of risk factors for non-communicable chronic dis- cally focus on the role that different versions of network eases. This situation is thus calling for prophylactic analysis, molecular signature identification and multi- strategies to promote healthy aging (Suzman et al. 2015). omics data integration are playing in elucidating the Aging is characterized by a progressive appearance of complex mechanisms underlying nutrition, and provide various dysfunctions that lead to physical and metabolic frailty, which increase the risk of developing various diseases (Lee et al. 2011). Notably, antioxidant and anti-in- This article is part of a Topical Collection in Genes and Nutrition on flammatory capacity are affected with increasing age as is ‘‘Systems Nutrition and Health’’, guest edited by Jim Kaput, Martin Kussmann and Marijana Radonjic. the release and accumulation of stressors which result in a state of chronic low-grade inflammation and oxidative Electronic supplementary material The online version of this stress conceptualized, respectively, as ‘‘inflamm-aging’’ article (doi:10.1007/s12263-015-0508-9) contains supplementary material, which is available to authorized users. and ‘‘the free radical theory of aging’’ (Rubinsztein et al. 2011; Baylis et al. 2013; Beekman et al. 2013; Cevenini & Laura Caberlotto et al. 2013; Castellani et al. 2015). Another hallmark of [email protected] aging is the progressive decline of the autophagy process, which normally plays a protective, anti-cytotoxic role by 1 The Microsoft Research, University of Trento Centre for Computational Systems Biology (COSBI), Piazza degrading protein aggregates and dysfunctional cellular Manifattura 1, 38068 Rovereto, Italy components including mitochondria (Rubinsztein et al. 2 Department of Mathematics, University of Trento, 2011). By doing so, autophagy prevents the release of pro- Via Sommarive 14, 38123 Povo, Italy oxidative and pro-inflammatory toxins, but also regulates 123 58 Page 2 of 10 Genes Nutr (2015) 10:58 inflammation by exerting a control over the NLRP3 mammalian models (Speakman and Mitchell 2011). It was inflammasome and interleukin 1b production (Rubinsztein postulated that CR benefits result from evolutionary et al. 2011). Protein synthesis rates are also decreased in mechanisms that halts cell division and energy-requiring the aging individuals, which, when combined with the functions under conditions of energy imbalance (Speakman often-inadequate physical activity and nutritional habits of and Mitchell 2011). The benefits of CR on slowing or this population, increases risk of sarcopenia with ensuing preventing age-related dysfunctions are the result of its metabolic dysfunctions (Beyer et al. 2012; Fabian et al. influence on the many systemic effectors previously men- 2012). tioned as being affected in aging. This is demonstrated by Age-related dysfunctions thus arise from several several studies which showed that CR (1) negatively reg- molecular systems that share central molecular effectors ulates the insulin-like growth factor (IGF-1)/insulin particularly in the sirtuin family (SIRTs, regulators of receptor substrate (IRS)/PI3k–Atk pathways, (2) activates cellular energy metabolism and mitochondrial biogenesis) SIRT family members, (3) activates AMPK (through and adenosine monophosphate-activated protein kinase decreased ATP/AMP ratio), (4) inhibits the mTOR path- (AMPK, central regulator of energy metabolism) and their way, (5) improves fatty acid metabolism (via CPT-1 and downstream targets (Salminen and Kaarniranta 2012; SREBP-1), (6) decreases the release of ROS and pro-in- Merksamer et al. 2013). Dysfunctions in those systems flammatory compounds and (7) modulates the expression principally result in decreased capacity to respond to of genes implicated in neuroprotection (Lee et al. 2000; stressors, to regulate mitochondrial function and energy Rubinsztein et al. 2011; Speakman and Mitchell 2011; Dai metabolism and, through their interaction with the mam- et al. 2014). malian target of rapamycin (mTOR), to impaired autop- These effectors have been shown to have organ-specific hagy mechanisms (Rubinsztein et al. 2011; Salminen and effects. They partially act in synergy to improve muscle Kaarniranta 2012; Merksamer et al. 2013). energy metabolism, prevent muscle loss and mitochondrial Altogether, these alterations affect energy homeostasis dysfunction in mice (Jang et al. 2012; Lin et al. 2014; Chen and increase the risk of pathologies that share common et al. 2015) and induce mitochondrial biogenesis in the inflammatory and oxidative basis such as obesity, insulin skeletal muscle (of human as well), the heart, the adipose resistance, type II diabetes and cardiovascular diseases tissue and the brain (Baur 2010). CR also activates (Ceriello et al. 2004; Scrivo et al. 2011; North and Sinclair autophagy (Rubinsztein et al. 2011; Speakman and 2012). In the aging brain, the excessive release of oxidative Mitchell 2011). In the aging rat brain, CR improved ketone and inflammatory mediators coupled with impaired body metabolism and blood flow, often impaired in neu- autophagy mechanisms result in microvascular dysfunc- rodegenerative disorders (Lin et al. 2015). CR also delayed tions, protein oxidation, lipid peroxidation and DNA neurodegeneration in aging mice through SIRT1 (Graff damage, which eventually lead to ischemic damages, et al. 2013), improved learning through PI3k/Akt pathway abnormal protein aggregation, neuronal inflammation and (Ma et al. 2014) and was demonstrated to attenuate amy- cell death common to a number of neurodegenerative dis- loid b neuropathology in aged mice with Alzheimer’s orders (Rubinsztein et al. 2011; Rege et al. 2014; Salminen disease (Wang et al. 2005). Finally, one of the few inves- and Paul 2014). Moreover, excessive oxidative stress alters tigation conducted in humans showed that a 3-month CR synaptic transmission and neuronal excitability, which intervention improved memory in the elderly (Witte et al. leads to structural damage of the central nervous system 2009). (CNS) and decline of cognitive functions (Rizzo et al. 2014; Salminen and Paul 2014). Calorie restriction mimetics: the case of polyphenols

Caloric restriction Caloric restriction, however, benefits from being initiated early in life to yield the most positive effects on aging, The pathophysiological implications of aging are thus particularly neuroprotective effects and was rarely inves- broad and include multiple systems. Promotion of healthy tigated in human partly because it is often accompanied by aging should therefore similarly have multiple pleiotropic side effects that make it hardly sustainable in the long-term targets of action. Approaches aimed at improving lifestyle (Speakman and Mitchell 2011). Alternative interventions and dietary habits are part of those strategies that have that would mimic—at least partially—the molecular and large spectrum or action. Caloric restriction (CR), referring physiological benefits of CR are thus of interest (Speakman to 20–40 % reduction in habitual daily energy intake, is and Mitchell 2011). Intermittent CR (or intermittent fast- one such intervention that has largely been investigated ing) is one such alternative previously shown to decrease since it was first identified to have potentially life- oxidative stress levels and improve insulin sensitivity in increasing benefits in yeast and, more recently, in humans, but more research has yet to be done in this area 123 Genes Nutr (2015) 10:58 Page 3 of 10 58

(Wegman et al. 2015). Polyphenols—a class of antioxidant biology approaches (Corthesy-Theulaz et al. 2005). These phytochemicals mostly found in plant-derived produce methods offer novel approaches to understanding the such as fruits and vegetables (and their derivatives such as complex molecular mechanisms by which nutrition juices and wines), nuts, spices and grains (tea and choco- affects—patho—physiological processes (e.g., aging) in late)—represent another nutritional alternative to CR. this increasingly data-rich field. Nutritional research has in In rats, resveratrol supplementation prevented age-in- fact gathered knowledge on the interaction of nutrients/ duced muscle loss (Jang et al. 2012; Joseph et al. 2013), nutritional habits with an individual’s genome or between suppressed pro-apoptotic signaling in senescent heart and gene polymorphisms and individual responses to nutrients/ prevented age-related heart dysfunction (Sin et al. 2014). nutritional habits. Moreover, data on how nutrition affects Moreover, supplementation with polyphenol-rich blueber- DNA methylation (nutritional epigenomics), gene expres- ries increased IGF-1 expression in hippocampus of aged sion (transcriptomics) and protein expression or post- rats, improving hippocampal plasticity and memory per- translational modifications (proteomics), and how these formances (Casadesus et al. 2004). In addition, resveratrol changes affect a large number of metabolites (metabo- supplementation in non-human primates increased working lomics) continue to be produced. As a final layer of com- (as did CR) and spatial memory performance (Dal-Pan plexity, gut microbiota composition (metagenomics) is et al. 2011). Furthermore, this class of compounds was increasingly recognized as crucial in modulating the two- shown to mimic the effects of CR at the transcriptomic way relationship between nutrition and host phenotypes. levels notably in reducing signs of aging (Barger et al. The recent, and ever-growing, availability of -omics 2008; Pearson et al. 2008; Vidal et al. 2011). In fact, data will widen the window of observation of biological polyphenols and resveratrol mediate multiple of the pre- and pathophysiological processes that often involve mul- viously described pathways such as SIRTs, AMPK and tiple systems from different cell and/or tissue types. To mTOR (Lam et al. 2013). In longitudinal clinical inter- fully profit from these advances, however, new method- ventions conducted in humans, increased consumption of ologies will need to be developed and deepened in order to polyphenol-rich and Mediterranean-type food items such as successfully integrate the information provided by the extra-virgin olive oil and nuts has been associated with different levels of -omics into comprehensive models in markers of healthy aging such as improvement of cardio- order to avoid the risk of drowning under the over-abun- vascular risk factors and cognition (Valls-Pedret et al. dance of such data. 2012; Martinez-Lapiscina et al. 2013; Zhu et al. 2014; The following sections provide an overview of such Houston et al. 2015). Moreover, a resveratrol supplemen- integrative approaches and will include an example on how tation study in healthy aged individuals improved brain these methods can be used to uncover new knowledge from functional connectivity, which correlated with improved existing microarray data produced to investigate the glucose metabolism and memory (Witte et al. 2014). molecular impact of CR and resveratrol supplementation on brain tissues of young and aged mice. Challenges facing novel -omics technologies

Nutritional interventions aimed at promoting healthy aging Systems biology influence multiple systems and have broad, and probably largely unknown, range of effectors. Furthermore, most Integrative approaches to studying biological systems are investigations were conducted in animal models or in aged often the focus of the emerging field of systems biology. cohorts in whom the optimal intervention window to obtain While a definitive and agreed-upon definition of what significant life span improvements might be passed. On the constitutes a systems biology method is still missing, it is other hand, longitudinal studies measuring hard endpoints generally accepted that the final objective is the study of such as decreases in aging-related morbidities or life span complex biological systems using a holistic approach and extension would require large investment of both time and relying on system-wide knowledge and high-throughput capital. Understanding of the mechanisms associated with data. Examples of the methods employed are network- healthy aging promoting interventions and identifying based analysis, methods based on molecular signatures, potent biomarkers able to monitor the early efficacy of multi-omics data integration and system-wide modeling. those interventions in younger cohorts are thus needed Herein, we will briefly review how the first three concepts (Belsky et al. 2015). Moreover, interindividual variability have been applied to the discovery of the connections in response to interventions adds to the necessity of iden- between nutrition and healthy aging. A specific case study tifying biomarkers of responders (Barberger-Gateau 2014). will be presented as an illustrative example, which includes Luckily, those needs can be fulfilled by recent high- a novel type of molecular signature combined with network throughput -omics capabilities and integrative systems analysis. We will not cover other important areas of 123 58 Page 4 of 10 Genes Nutr (2015) 10:58 systems biology that would be best reviewed separately, closely overlap with known biological functions (Ravasz such as the field of multi-scale modeling of biological et al. 2002; Parikshak et al. 2015), thus increasing the systems, that also play a prominent role in the research on effectiveness of the functional analysis step. aging (see for example Milanesi et al. 2009). A limited number of studies in nutritional research have used systems biology approaches. For example, seeded approach to network analysis was applied by Wuttke et al. Network-based analysis (2012). The authors started with a list of CR-related genes (used as seeds) compiled from the literature and expanded Complex molecular systems such as gene regulation, pro- it by addition of their highly interacting neighbors in order tein–protein interactions (PPI) or the metabolic machinery, to predict novel CR-influenced genes in different organ- can be conveniently represented as networks. In such a isms. Using the expanded gene list, they then performed a network representation, molecules are shown as nodes, and large-scale comparison of datasets (interactomes and tran- the relationship between molecules as edges. Relationships scriptomes) from multiple organisms. This enabled them to can be correlations between gene expression levels, PPIs, identify the most essential elements as those most con- and metabolic interactions coming from experimental served between organisms and conclude that life span studies (and derived from statistical associations) or data- extension operates via ancient rejuvenation process derived bases, or curated from the literature. from gametogenesis. The study of biological networks and their modeling, The work of Morine et al. is an example of genome- analysis and visualization represent powerful and flexible wide approach to network analysis in which the authors tools that can be useful in gathering data produced by high- studied the transcriptomic signature of adipose tissue of throughput techniques and in helping uncover the knowl- obese IL-1RI-/- mice subjected to high-fat feeding edge thus generated. Information used to build networks (Morine et al. 2013). Following intervention, markers of can be qualitative or quantitative (unweighted vs weighted insulin sensitivity and inflammation in adipose tissue were interactions), causal or correlative (directed vs undirected assessed with the aid of a PPI/regulatory network of innate edges), qualified according to strength of evidence (inter- immunity and identified a significantly enriched module actions with qualifying labels such as ‘‘experimental,’’ highlighting potentially novel inflammatory mediators of ‘‘literature,’’ ‘‘indirect,’’ ‘‘correlation-based’’). Further- adipogenesis. more, such networks can be curated to specific cell type, These studies demonstrated how network analysis can tissue or condition by, for example, only considering genes clarify the multi-systemic impacts of nutrition and how it or proteins ubiquitously expressed/present in all tissues and could be applied to further the understanding of the in the tissue of interest (Tegner et al. 2007). mechanisms underlying healthy aging promoting Two approaches can be used to analyze biological net- interventions. works, namely the seeded and genome-wide approaches (Parikshak et al. 2015). The seeded approach uses biomolecules of interest (e.g., obtained from experimental or Molecular signatures genetic analysis) typically in the form of a set of differentially expressed genes or proteins, as the basis to build a Another powerful tool for the study of -omics data is the network by addition of highly connected neighbors of those concept of molecular signatures. Network analysis and seeds. Upon optional network module detection, functional signature analysis represent complementary approaches to annotation analysis can then be carried out using seeds and the study of complex systems. While network analysis their highly connected neighbors as input. This method as seeks to capture and manage the complexity of a biological the advantage of adding information missed by the scarcity system, signature analysis summarizes the essential fea- of biomolecules experimentally quantified and can poten- tures of such a system in the most possible succinct way. tially increase the significance of the results of a traditional The simplest form of molecular signature is a list of bio- functional analysis such as GO enrichment analysis. molecules (genes, proteins or metabolites) carefully The genome-wide approach is a slightly more advanced selected for their discriminating value; such signatures version in that it identifies modules of highly connected have long been studied for their potential as diagnostic nodes within a comprehensive genome-wide network. biomarkers. Since our understanding of the effects of Biomolecules of interest are then mapped onto the genome- nutrition on aging in humans would greatly benefit from an wide network, and modules significantly enriched with effective form of monitoring, the need for biomarkers in biomolecules of interest are identified and analyzed for this area of research is a pressing one. Traditional bio- functional enrichment. The advantage of this approach is marker identification identified a set of genes rapidly that modules in molecular networks have been shown to affected upon initiation of CR in old mice (Dhahbi et al. 123 Genes Nutr (2015) 10:58 Page 5 of 10 58

2004). Those genes could thus be good candidate robustness to batch effects and to differences in processing biomarkers to rapidly assess the efficacy of, and detect methods, and the possibility of using expression data from responders from non-responders to CR. multiple sources, none of which were relevant for this particular study. We re-analyzed the microarray data produced by Barger et al. (2008) on the role of CR and Enhanced network-based analysis: combining resveratrol at low doses on brain aging. Briefly, 14-month- networks and signatures old male mice fed a control diet were randomly assigned to one of the following: a CR (63 kcal weekly) diet, a Besides their use as biomarkers, molecular signatures can resveratrol-supplemented (50 mg resveratrol per kilogram also be used as a selection tool to increase effectiveness of of body weight) control diet and a control diet. Brain tis- network analysis. Network analysis and signature extrac- sues were collected at 30 month, while brain tissues from tion can be combined in a single workflow in at least two 5-month-old mice fed the control diet were also collected possible ways, which we term the ‘‘summary’’ and the and served as young controls. Identification of the tran- ‘‘feature-list’’ approaches. In the summary approach, the scriptional signature was done using an enhanced version signature extraction operation follows the network analysis of a rank-based classification method described in the and is performed on the results of the latter. In this case, Supplementary Material (Lauria 2013; Tarca et al. 2013; signatures are typically used as a succinct way to sum- Lauria et al. 2015). marize the complex outcome of the network analysis as a Following a version of the seeded approach to network way to facilitate, for example, the comparison of results analysis, a network was then build using the signature across treatments, time points, tissues or conditions. The genes as seeds with the NetWalker tool (Komurov et al. advantage of the summary approach is that it results in a 2012) in combination with the BioGrid PPI network for list of biomolecules that are more relevant to the question mus musculus (rel. 3.2.121). The resulting NetWalker net- of interest than a signature obtained directly from the - work, thus, contained a subset of highly connected signa- omics data, because of the discriminating power of the ture genes (i.e., those interacting above a determined intervening network analysis (Kelder et al. 2014). A sum- threshold), augmented by additional genes that were con- mary approach was selected to study the hepatic molecular nected to the signature ones through above-threshold network underlying type 2 diabetes mellitus using tran- interactions. The genes obtained from such network scriptional data obtained from experiments with different enrichment analysis were then used to extract the most drug and dietary treatments (Kelder et al. 2014). After representative GO biological process terms (i.e., the ones performing a genome-wide network analysis, the authors that are over-represented, but that do not refer to most compared signatures of the different interventions, which general biological processes). Pathway analysis was per- highlighted functionally related nodes that may be used as formed using DAVID web-based tool using hypergeomet- target to future anti-diabetic treatments or as biomarkers ric distribution test and a FDR-adjusted p value threshold for determining their efficacy (Fig. 1). of 0.05 (Huang da et al. 2009). In the feature-list approach, all -omics data are used as Using this method, we identified transcriptomic signa- input for the signature extraction operation, and the bio- tures and related networks for age (5- vs 30-month-old molecules included in the signatures are then used as input mice), CR (30 month control vs 30 month CR) and to network analysis (e.g., as the seeds in case of the seeded resveratrol (30 month control vs 30 month resveratrol). approach). The advantage of this approach is that the In the brain aging process (young vs old mice on control ensuing network analysis inherits all the potential benefits diet), the analysis of the Netwalker-enriched network (see of the selected signature extraction algorithm, possibly Fig. 2) highlighted biological functions associated with overcoming the known limitations of traditional statistical immune processes and inflammation, steroid hormone methods for the identification of differently expressed receptor signaling, neurogenesis and histone modification genes. (Supplementary Table 1). Caloric restriction and resvera- As a demonstration of the feature-list approach, we trol supplementation produced overlapping results related apply our recently developed signature identification to neuronal functions including regulation of synaptic method followed by network analysis, in order to study the transmission and glutamate receptor signaling. In addition, role of CR and resveratrol supplementation on brain aging GO BP term associated with response to oxidative stress in mice (Fig. 2). The novelty represented by the use of a was significantly enriched in resveratrol-treated mice and transcriptional signature enabled a more sensitive analysis with a trend toward significance in CR-exposed mice, than the one resulting from a list of differentially expressed while inflammatory processes were exclusively associated genes obtained with traditional methods. Other benefits with resveratrol supplementation. Overlapping GO BP afforded by our signature identification method are terms between both dietary interventions demonstrate an 123 58 Page 6 of 10 Genes Nutr (2015) 10:58

Fig. 1 Illustration of the two main approaches to network analysis, namely the seeded (left) and genome-wide (right)

Fig. 2 Schematic representation of the case study workflow. Tran- each group. Network analysis was then performed using Netwalker scriptomic data from brain tissue of treatment groups (caloric method and the mouse-specific BioGRID PPI database. Finally, the restriction, low-dose resveratrol supplementation and control diet) functional enrichments of the genes obtained from the network from young and aged mice were analyzed using a novel approach analysis were characterized by investigating over-represented gene enabling the identification of transcriptional signature characterizing ontology biological processes and pathways effect on synaptic transmission particularly associated with Complete lists of Materials and methods and results are ionotropic glutamate receptors, while only resveratrol available in the supplementary material. supplementation had a significant effect on NF-kB Sig- Although these findings largely confirm those of Barger naling (Supplementary Table 1). et al. and are in agreement with previous knowledge (Lam

123 Genes Nutr (2015) 10:58 Page 7 of 10 58 et al. 2013; Marchal et al. 2013), our method was also able the -omics data available. For example, -omics integration to uncover the effect of both interventions on unrecognized can be used to understand the genetic architecture of molecular processes such as those related to synaptic complex traits by integrating genomics with transcrip- transmission and glutamatergic signaling. In particular, tomics, proteomics, and/or metabolomics, to evaluate the they suggest the involvement of glutamatergic ionotropic role of genetic variations in intermediate molecular phe- receptors, including AMPA receptor subunits 1, 2 and 3 notype of a complex trait. Such studies are called expres- (GRIA1, 2 and 3), a result reported in connection with CR sion QTL (eQTL; Schadt 2006), proteomics QTL (pQTL) in rats (Shi et al. 2007) and very recently confirmed also in or metabolomics QTL (mQTL) analysis depending on the mice (Schafer et al. 2015). The involvement of these key data available. Multi-omics analysis can also be used to signaling molecules for brain functions reinforces the obtain a comprehensive modeling of -omics profiles to be notion that dietary interventions such as those studied used as biomarker of disease, disease state and/or treatment herein hold the ability to modulate brain health and func- efficacy, or to determine key elements, which correlate tion. This is of particular interest for resveratrol supple- with a specific phenotype, or for exploring host–microbiota mentation as knowledge was limited on its ability to interactions. modulate glutamatergic transmission and further supports There are two main strategies for integration: multistage the investigation of resveratrol in the treatment of various (stepwise or sequential) and meta-dimensional (simulta- aging-related central nervous system disorders. Interest- neous) analyses (Ritchie et al. 2015). The multistage ingly, this resveratrol effect has been confirmed indepen- analyses are stepwise or hierarchical methods that rely on dently in rats (Wang et al. 2015). the central dogma of biology in which variations at the Moreover, our analyses confirm the effects of age on DNA level will hierarchically affect RNA levels, protein NF-kB signaling and inflammation, particularly on Toll- expression and so on. These methods reduce the search like receptor (TLR) signaling, but also interestingly high- space through different steps of analysis but are challenged light impacts of aging on the regulation of neurogenesis, by the accumulating evidence (e.g., notably epigenetic known to decline with advancing age (Kuhn et al. 1996; modifications, microbiota–host interactions) that question Villeda et al. 2011). Among the age-related plasma factors the linearity of such dogma and thus the ability of this correlated strongly with decreased neurogenesis, b2 model to capture every inter-omics interaction. microglobulin was found also in our age signature and in The more computationally demanding meta-dimen- the Netwalker-enriched network, suggesting that sional analyses simultaneously combine multi-omics data immunological response could be related to age-related to produce complex models defined by multiple differently impairments of neurogenesis (Villeda et al. 2011). scaled variables. Those approaches are of interest since This re-examination of previously published data shows every potential relationship between different -omics levels that a novel transcriptional signature identification com- are considered and, thus, it captures potential inter-omics bined with carefully matched network analysis can provide interactions more realistically. new insight into the mechanisms involved in the complex Multi-omics integration can be achieved simply by interaction of nutritional interventions and aging and guide separately analyzing different -omic data and then com- potential new therapeutic approaches. bining the obtained results together in order to draw global conclusions. By doing so, this methodology does not, however, provide statistics on the interactions between - Multi-omics data integration omics levels since they were treated separately. Statistical integration methods perform statistical association between Traditional and network-based analysis of the major plat- biomolecules from different -omics levels (Cavill et al. forms of comprehensive -omics can be used to study the 2015; Ritchie et al. 2015). The main statistical integration influence of nutritional interventions (for instance CR and approaches can either be based on correlation, concatena- polyphenol supplementation) on biological processes such tion, multivariate or pathway analysis, which are further as those involved in aging. However, previously stated detailed in Table 1 along with suggestions of useful tools investigations assessed biological processes/dysfunctions that implement each approach. on the basis of data representative of single-omics level. To our knowledge, there are no examples of multi-omics Systems biology would benefit from a holistic system-wide integration analysis applied in the field of CR or approach by integrating different -omics levels with the polyphenol supplementation on age-related pathologies. following methodologies and tools. Although relatively sparse, there are nonetheless some Many approaches to perform multi-omics data integra- examples of studies that used those approaches to analyze tion have been developed and have different usefulness the effect of dietary interventions on other phenotypes. For depending on the biological question to be addressed and instance, Montastier and collaborators have recently 123 58 Page 8 of 10 Genes Nutr (2015) 10:58

Table 1 Details of the four main statistical integration approaches and suggested tools Method Comments Suggested tools Tool ref.

Correlation- Seeks to identify correlative links between elements of 3Omics Kuo et al. (2013) based different -omics datasets integration Concatenation- Groups different -omics or single-omic dataset from Multiple co-inertia analysis (MCIA) Meng et al. (2014) based different experiments or experimental conditions into a Joint and Individual Variation Explained Lock et al. (2013) integration single matrix and then performs integrated analysis (JIVE) Multivariate- Relies on multivariate statistical methods such as canonical R package mixOmics Gonza´lez et al. based correlation analysis (CCA) (Jozefczuk et al. 2010) and (2011) integration orthogonal partial least squares discriminant analysis (OPLS-DA) (Boccard and Rutledge 2013) to calculate relationship between different levels of -omics data Pathway-based Integrates different -omics levels by relying on existing InCroMAP and IMPALA for integrated Eichner et al. integration biological knowledge gathered from metabolic pathways pathway-based analysis (2014), such as Kegg and wikipathways (Kutmon et al. 2015) Kamburov et al. (2011) SAMNetWeb to generate biological Gosline et al. networks with transcriptomics and (2015) proteomics data MetScape cytoscape plugin to produce Karnovsky et al. metabolic networks from transcriptomics (2012) and metabolite data performed a multistage analysis using both single and field of nutrition. This review has illustrated the potential multi-omics networks to show metabolic alterations of network analysis, molecular signature identification and occurring during weight change in response to CR (Mon- multi-omics data integration to generate candidate tastier et al. 2015). To do so, they first inferred a partial biomarkers and novel molecular mechanisms in an unbi- correlation network for each -omics level and then con- ased fashion. We foresee that the real potential for nutri- structed multi-omics networks linking each pairs of data tional systems biology applications will lie in multi- type using regularized canonical correlation analysis, knowledge integration strategies that will, by including which successfully infers a gene/phenotype network information at different levels, shed light on gene–nutrient (Rengel et al. 2012). They then merged single and multi- interactions with a degree of accuracy and completeness omics networks together and performed module detection not yet achievable today. analyses. Temporal analysis of the modules successfully revealed both shared and time-specific biological signatures in response to metabolic variations occurring during References weight changes. Finally, multi-omics integration that would take into Barberger-Gateau P (2014) Nutrition and brain aging: how can we move ahead? Eur J Clin Nutr 68:1245–1249 account cell or tissue types, as proposed by Tegner and col- Barger JL, Kayo T et al (2008) A low dose of dietary resveratrol laborators, would be desirable to investigate the cell- or tissue- partially mimics caloric restriction and retards aging parameters specific effects of pleiotropic interventions such as nutri- in mice. PLoS ONE 3:e2264 tional—and lifestyle—modification modulating progression Baur JA (2010) Resveratrol, sirtuins, and the promise of a DR mimetic. Mech Ageing Dev 131:261–269 of complex, multi-system diseases (Tegner et al. 2007). Baylis D, Bartlett DB et al (2013) Understanding how we age: insights into inflammaging. Longev Healthspan 2:8 Beekman M, Blanche H et al (2013) Genome-wide linkage analysis Conclusion for human longevity: Genetics of Healthy Aging Study. Aging Cell 12:184–193 Belsky DW, Caspi A et al (2015) Quantification of biological aging in There is a pressing need for new methods to tackle the young adults. Proc Natl Acad Sci USA 112(30):E4104–E4110. complexity of the interplay between nutritional interven- doi:10.1073/pnas.1506264112 tions and aging and to make sense of a growing amount of - Beyer I, Mets T et al (2012) Chronic low-grade inflammation and age- related sarcopenia. Curr Opin Clin Nutr Metab Care 15:12–22 omics data being produced for this purpose. In this review, Boccard J, Rutledge DN (2013) A consensus orthogonal partial least we detailed systems biology-inspired methods of analysis squares discriminant analysis (OPLS-DA) strategy for multi- that could fulfill this need and be applied to the burgeoning block Omics data fusion. Anal Chim Acta 769:30–39

123 Genes Nutr (2015) 10:58 Page 9 of 10 58

Casadesus G, Shukitt-Hale B et al (2004) Modulation of hippocampal Kelder T, Verschuren L et al (2014) Network signatures link hepatic plasticity and cognitive behavior by short-term blueberry effects of anti-diabetic interventions with systemic disease supplementation in aged rats. Nutr Neurosci 7:309–316 parameters. BMC Syst Biol 8:108 Castellani GC, Menichetti G et al (2015) Systems medicine of Komurov K, Dursun S et al (2012) NetWalker: a contextual network inflammaging. Brief Bioinform 1–14. doi:10.1093/bib/bbv062 analysis tool for functional genomics. BMC Genom 13:282 Cavill R, Jennen D et al (2015) Transcriptomic and metabolomic data Kuhn HG, Dickinson-Anson H et al (1996) Neurogenesis in the integration. Brief Bioinform 1–11. doi:10.1093/bib/bbv090 dentate gyrus of the adult rat: age-related decrease of neuronal Ceriello A, Motz E (2004) Is oxidative stress the pathogenic progenitor proliferation. J Neurosci 16:2027–2033 mechanism underlying insulin resistance, diabetes, and cardio- Kuo TC, Tian TF et al (2013) 3Omics: a web-based systems biology vascular disease? The common soil hypothesis revisited. Arte- tool for analysis, integration and visualization of human transcrip- rioscler Thromb Vasc Biol 24:816–823 tomic, proteomic and metabolomic data. BMC Syst Biol 7:64 Cevenini E, Monti D et al (2013) Inflamm-ageing. Curr Opin Clin Kutmon M, Riutta A et al (2015) WikiPathways: capturing the full Nutr Metab Care 16:14–20 diversity of pathway knowledge. Nucleic Acids Res. doi:10. Chen CN, Lin SY et al (2015) Late-onset caloric restriction alters 1093/nar/gkv1024 skeletal muscle metabolism by modulating pyruvate metabolism. Lam YY, Peterson CM et al (2013) Resveratrol vs. calorie restriction: Am J Physiol Endocrinol Metab 308:E942–E949 data from rodents to humans. Exp Gerontol 48:1018–1024 Corthesy-Theulaz I, den Dunnen JT et al (2005) Nutrigenomics: the Lauria M (2013) Rank-based transcriptional signatures: a novel impact of biomics technology on nutrition research. Ann Nutr approach to diagnostic biomarker definition and analysis. Syst Metab 49:355–365 Biomed 1:228–239 da Huang W, Sherman BT et al (2009) Bioinformatics enrichment Lauria M, Moyseos P et al (2015) SCUDO: a tool for signature-based tools: paths toward the comprehensive functional analysis of clustering of expression profiles. Nucleic Acids Res 43:W188– large gene lists. Nucleic Acids Res 37:1–13 W192 Dai DF, Karunadharma PP et al (2014) Altered proteome turnover Lee CK, Weindruch R et al (2000) Gene-expression profile of the and remodeling by short-term caloric restriction or rapamycin ageing brain in mice. Nat Genet 25:294–297 rejuvenate the aging heart. Aging Cell 13:529–539 Lee JS, Auyeung TW et al (2011) Physical frailty in older adults is Dal-Pan A, Pifferi F et al (2011) Cognitive performances are associated with metabolic and atherosclerotic risk factors and selectively enhanced during chronic caloric restriction or cognitive impairment independent of muscle mass. J Nutr Health resveratrol supplementation in a primate. PLoS ONE 6:e16581 Aging 15:857–862 Dhahbi JM, Kim HJ et al (2004) Temporal linkage between the Lin AL, Coman D et al (2014) Caloric restriction impedes age-related phenotypic and genomic responses to caloric restriction. Proc decline of mitochondrial function and neuronal activity. J Cereb Natl Acad Sci USA 101:5524–5529 Blood Flow Metab 34:1440–1443 Eichner J, Rosenbaum L et al (2014) Integrated enrichment analysis Lin AL, Zhang W et al (2015) Caloric restriction increases ketone and pathway-centered visualization of metabolomics, pro- bodies metabolism and preserves blood flow in aging brain. teomics, transcriptomics, and genomics data by using the Neurobiol Aging 36:2296–2303 InCroMAP software. J Chromatogr B Analyt Technol Biomed Lock EF, Hoadley KA et al (2013) Joint and individual variation Life Sci 966:77–82 explained (Jive) for integrated analysis of multiple data types. Fabian E, Bogner M et al (2012) Vitamin status in elderly people in Ann Appl Stat 7:523–542 relation to the use of nutritional supplements. J Nutr Health Ma L, Zhao Z et al (2014) Caloric restriction can improve learning Aging 16:206–212 ability in C57/BL mice via regulation of the insulin-PI3K/Akt Gonza´lez I, Le´ Cao K et al (2011) mixOmics: Omics Data Integration signaling pathway. Neurol Sci 35:1381–1386 Project. http://www.mixomics.org Marchal J, Dal-Pan A et al (2013) Calorie restriction and resveratrol Gosline SJ, Oh C et al (2015) SAMNetWeb: identifying condition- supplementation prevent age-related DNA and RNA oxidative specific networks linking signaling and transcription. Bioinfor- damage in a non-human primate. Exp Gerontol 48:992–1000 matics 31:1124–1126 Martinez-Lapiscina EH, Clavero P et al (2013) Virgin olive oil Graff J, Kahn M et al (2013) A dietary regimen of caloric restriction supplementation and long-term cognition: the PREDIMED- or pharmacological activation of SIRT1 to delay the onset of NAVARRA randomized, trial. J Nutr Health Aging 17:544–552 neurodegeneration. J Neurosci 33:8951–8960 Meng C, Kuster B et al (2014) A multivariate approach to the Houston DK, Leng X et al (2015) A long-term intensive lifestyle integration of multi-omics datasets. BMC Bioinformatics 15:162 intervention and physical function: the look AHEAD Movement Merksamer PI, Liu Y et al (2013) The sirtuins, oxidative stress and and Memory Study. Obesity (Silver Spring) 23:77–84 aging: an emerging link. Aging (Albany NY) 5:144–150 Jang YC, Liu Y et al (2012) Dietary restriction attenuates age- Milanesi L, Romano P et al (2009) Trends in modeling biomedical associated muscle atrophy by lowering oxidative stress in mice complex systems. BMC Bioinform 10(Suppl 12):I1 even in complete absence of CuZnSOD. Aging Cell 11:770–782 Montastier E, Villa-Vialaneix N et al (2015) System model network Joseph AM, Malamo AG et al (2013) Short-term caloric restriction, for adipose tissue signatures related to weight changes in resveratrol, or combined treatment regimens initiated in late-life response to calorie restriction and subsequent weight mainte- alter mitochondrial protein expression profiles in a fiber-type nance. PLoS Comput Biol 11:e1004047 specific manner in aged animals. Exp Gerontol 48:858–868 Morine MJ, Toomey S et al (2013) Network analysis of adipose tissue Jozefczuk S, Klie S et al (2010) Metabolomic and transcriptomic gene expression highlights altered metabolic and regulatory stress response of Escherichia coli. Mol Syst Biol 6:364 transcriptomic activity in high-fat-diet-fed IL-1RI knockout Kamburov A, Cavill R et al (2011) Integrated pathway-level analysis mice. J Nutr Biochem 24:788–795 of transcriptomics and metabolomics data with IMPaLA. Bioin- North BJ, Sinclair DA (2012) The intersection between aging and formatics 27:2917–2918 cardiovascular disease. Circ Res 110:1097–1108 Karnovsky A, Weymouth T et al (2012) Metscape 2 bioinformatics Parikshak NN, Gandal MJ et al (2015) Systems biology and gene tool for the analysis and visualization of metabolomics and gene networks in neurodevelopmental and neurodegenerative disor- expression data. Bioinformatics 28:373–380 ders. Nat Rev Genet 16:441–458

123 58 Page 10 of 10 Genes Nutr (2015) 10:58

Pearson KJ, Baur JA et al (2008) Resveratrol delays age-related Suzman R, Beard JR et al (2015) Health in an ageing world—what do deterioration and mimics transcriptional aspects of dietary we know? Lancet 385:484–486 restriction without extending life span. Cell Metab 8:157–168 Tarca AL, Lauria M et al (2013) Strengths and limitations of Ravasz E, Somera AL et al (2002) Hierarchical organization of microarray-based phenotype prediction: lessons learned from the modularity in metabolic networks. Science 297:1551–1555 IMPROVER Diagnostic Signature Challenge. Bioinformatics Rege SD, Geetha T et al (2014) Neuroprotective effects of resveratrol 29:2892–2899 in Alzheimer disease pathology. Front Aging Neurosci 6:218 Tegner J, Skogsberg J et al (2007) Thematic review series: systems Rengel D, Arribat S et al (2012) A gene-phenotype network based on biology approaches to metabolic and cardiovascular disorders. genetic variability for drought responses reveals key physiolog- Multi-organ whole-genome measurements and reverse engineer- ical processes in controlled and natural environments. PLoS ing to uncover gene networks underlying complex traits. J Lipid ONE 7:e45249 Res 48:267–277 Ritchie MD, Holzinger ER et al (2015) Methods of integrating data to Valls-Pedret C, Lamuela-Raventos RM et al (2012) Polyphenol-rich uncover genotype-phenotype interactions. Nat Rev Genet foods in the Mediterranean diet are associated with better 16:85–97 cognitive function in elderly subjects at high cardiovascular risk. Rizzo V, Richman J et al (2014) Dissecting mechanisms of brain J Alzheimers Dis 29:773–782 aging by studying the intrinsic excitability of neurons. Front Vidal M, Cusick ME et al (2011) Interactome networks and human Aging Neurosci 6:337 disease. Cell 144:986–998 Rubinsztein DC, Marino G et al (2011) Autophagy and aging. Cell Villeda SA, Luo J et al (2011) The ageing systemic milieu negatively 146:682–695 regulates neurogenesis and cognitive function. Nature 477:90–94 Salminen A, Kaarniranta K (2012) AMP-activated protein kinase Wang J, Ho L et al (2005) Caloric restriction attenuates beta-amyloid (AMPK) controls the aging process via an integrated signaling neuropathology in a mouse model of Alzheimer’s disease. network. Ageing Res Rev 11:230–241 FASEB J 19:659–661 Salminen LE, Paul RH (2014) Oxidative stress and genetic markers of Wang G, Amato S et al (2015) Resveratrol up-regulates AMPA suboptimal antioxidant defense in the aging brain: a theoretical receptor expression via AMP-activated protein kinase-mediated review. Rev Neurosci 25:805–819 protein translation. Neuropharmacology 95:144–153 Schadt EE (2006) Novel integrative genomics strategies to identify Wegman MP, Guo MH et al (2015) Practicality of intermittent fasting genes for complex traits. Anim Genet 37(Suppl 1):18–23 in humans and its effect on oxidative stress and genes related to Schafer MJ, Dolgalev I et al (2015) Calorie restriction suppresses age- aging and metabolism. Rejuvenation Res 18:162–172 dependent hippocampal transcriptional signatures. PLoS ONE Witte AV, Fobker M et al (2009) Caloric restriction improves 10:e0133923 memory in elderly humans. Proc Natl Acad Sci USA Scrivo R, Vasile M et al (2011) Inﬂammation as ‘‘common soil’’ of 106:1255–1260 the multifactorial diseases. Autoimmun Rev 10:369–374 Witte AV, Kerti L et al (2014) Effects of resveratrol on memory Shi L, Adams MM et al (2007) Caloric restriction eliminates the performance, hippocampal functional connectivity, and glucose aging-related decline in NMDA and AMPA receptor subunits in metabolism in healthy older adults. J Neurosci 34:7862–7870 the rat hippocampus and induces homeostasis. Exp Neurol Wuttke D, Connor R et al (2012) Dissecting the gene network of 206:70–79 dietary restriction to identify evolutionarily conserved pathways Sin TK, Yu AP et al (2014) Modulating effect of SIRT1 activation and new functional genes. PLoS Genet 8:e1002834 induced by resveratrol on Foxo1-associated apoptotic signalling Zhu N, Jacobs DR Jr et al (2014) Cardiorespiratory ﬁtness and in senescent heart. J Physiol 592:2535–2548 cognitive function in middle age: the CARDIA study. Neurology Speakman JR, Mitchell SE (2011) Caloric restriction. Mol Aspects 82:1339–1346 Med 32:159–221

123